Poster
One-Sided Unsupervised Domain Mapping
Sagie Benaim · Lior Wolf
Pacific Ballroom #92
Keywords: [ Unsupervised Learning ] [ Computer Vision ] [ Adversarial Networks ]
[
Abstract
]
Abstract:
In unsupervised domain mapping, the learner is given two unmatched datasets AA and BB. The goal is to learn a mapping GABGAB that translates a sample in AA to the analog sample in BB. Recent approaches have shown that when learning simultaneously both GABGAB and the inverse mapping GBAGBA, convincing mappings are obtained. In this work, we present a method of learning GABGAB without learning GBAGBA. This is done by learning a mapping that maintains the distance between a pair of samples. Moreover, good mappings are obtained, even by maintaining the distance between different parts of the same sample before and after mapping. We present experimental results that the new method not only allows for one sided mapping learning, but also leads to preferable numerical results over the existing circularity-based constraint. Our entire code is made publicly available at~\url{https://github.com/sagiebenaim/DistanceGAN}.
Live content is unavailable. Log in and register to view live content